|Publication number||US8026950 B2|
|Application number||US 10/570,427|
|Publication date||Sep 27, 2011|
|Filing date||Sep 3, 2004|
|Priority date||Sep 4, 2003|
|Also published as||US20070165129, WO2005025239A1|
|Publication number||10570427, 570427, PCT/2004/51679, PCT/IB/2004/051679, PCT/IB/2004/51679, PCT/IB/4/051679, PCT/IB/4/51679, PCT/IB2004/051679, PCT/IB2004/51679, PCT/IB2004051679, PCT/IB200451679, PCT/IB4/051679, PCT/IB4/51679, PCT/IB4051679, PCT/IB451679, US 8026950 B2, US 8026950B2, US-B2-8026950, US8026950 B2, US8026950B2|
|Inventors||Lyndon Hill, Graham R. Jones|
|Original Assignee||Sharp Kabushiki Kaisha|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (32), Non-Patent Citations (2), Referenced by (10), Classifications (17), Legal Events (2)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The present invention relates to a method of and an apparatus for selecting a stereoscopic pair of images. The present invention also relates to computer programs for such methods. Such techniques may be used, for example, in photography, videography, movie production and any application where stereoscopic images may be required. Such techniques may be used at the time of capturing stereoscopic images or may be used following capture, for example, to select stereoscopic pairs from still images or video sequences.
Three dimensional images as used in most 3D display technologies, e.g. EP 0 602 934, EP 0 656 555, EP 0 708 351, EP 0 726 482, EP 0 829 743 and EP 0 860 728 are formed by two 2D images. These displays direct separate images to either eye so that one eye sees one image and the other eye sees the other image. The Human Visual System (HVS) internally fuses the two images into a 3D image so that a person may sense depth from the displayed images.
In order to make stereoscopic images which may be viewed comfortably, the two images must be well aligned. This can be accomplished either by very accurate placement of the camera before capturing the images or by a post processing stage such as image rectification, for example as disclosed in EP 1 235 439. Image rectification is a method for re-aligning the two images so that they are parallel in three dimensions. This involves reprojecting them onto a common parallel plane by a geometrical transformation. When the images are parallel in this manner, they are considered to be rectilinear. If the images are not rectilinear, then the stereo image will be uncomfortable to view due to vertical disparities, perspective effects, excessive depth, etc, that the HVS cannot reconcile and the 3D effect breaks down.
Image rectification methods usually use powerful computer vision algorithms. However if the initial pair of images are badly aligned, such algorithms will be very slow or may completely fail. If a large amount of rectification is required, then the image quality may suffer when the two images are reprojected. Furthermore, the image overlap between the rectified images may be reduced and less usable image area will be available for the 3D image. Conversely, if the initial image alignment is good, the rectification can be processed much faster and undesirable side effects from image reprojection will be reduced. Therefore, if the two initial images are taken from well aligned positions, then a more comfortable 3D image may be created and a minimal amount of rectification will be required.
2001-230955 discloses a technique which is used in the Pentax Optio 230 digital camera and which provides a user-guided two shot stereo photography mode. In this method, the user takes the first image and is advised to “move the camera to the right”. A transparent vertical or horizontal strip of the first image is superimposed on the live image. The user is meant to align the second image to the first image using this strip. However, when this technique is followed, the camera will be rotated relative to its orientation when the first image was taken so that perspective errors are introduced and the two images are not rectilinear, resulting in a 3D image of poor visual quality.
EP1085769 discloses a variable separation dual-head camera that may utilise a method of image rectification such as that disclosed in EP1089573 for determining the amount of separation. This system is based on a priori knowledge of the scene. There is currently no known automated system for providing a recommendation of the separation required for producing 3D images which may be comfortably viewed.
The term “image feature” as used hereinafter is defined to mean anything which appears in an image. This includes, for example, lines, patterns and shading. This term is being used with its conventional meaning in the technical field of vision research.
According to a first aspect of the invention, there is provided a method of selecting a stereoscopic pair of images, comprising the steps of:
(a) selecting a first image;
(b) aligning at least one cursor overlaid on the first image with at least one image feature of the first image;
(c) shifting the or each cursor laterally by a predetermined amount; and
(d) selecting a second image which contains the feature and in which the or each cursor is overlaid on the second image and is substantially aligned with the or each feature.
The predetermined amount may correspond substantially to an average inter-ocular separation.
The predetermined amount may correspond to at least one parameter of a display for displaying the first and second images.
The predetermined amount may be dependent on image feature depth in the first image.
The predetermined amount may correspond to a maximum comfortable viewing depth.
The step (b) may comprise aligning the or each cursor manually.
The steps (a) and (d) may comprise selecting the first and second images from a set of precaptured images.
The step (a) may comprise capturing the first image by means of a camera and, the step (d) may comprise capturing the second image by means of the camera when the or each cursor is substantially aligned with the or each feature. The second image may be captured manually. The step (a) may be performed between the steps (b) and (c). The camera may have an optical axis and may be oriented such that the optical axis when the second image is captured is substantially parallel to the optical axis when the first image is captured. The positions of the camera from which the first and second images are captured may be spaced from each other substantially laterally and substantially perpendicularly to the optical axis when the first image was captured.
The or each cursor may comprise at least one of a grid of orthogonal lines, a cross-hair, a plurality of dots, a symbol defining a region of non-zero area, and a portion of the first image. The symbol may be a substantially rectangular outline. The cursor may be displayed alongside a laterally shifted cursor in both the first and second images.
The step (d) may comprise overlaying part of the first image on the second image. The part of the first image may be modified before overlaying, for example by being spatially highpass filtered. The part of the first image may comprise the portion of the first image overlaid by the region.
The step (d) may comprise performing a correlation between the first image and at least one candidate second image. The step (d) may comprise the user selecting the second image based on an indication of the level of correlation. The step (d) may comprise selecting the second image when the correlation exceeds a threshold. The step (d) may comprise selecting the second image when the correlation is a maximum.
The at least one cursor may comprise the symbol and the correlation may be performed between a part of the first image overlaid by the symbol and a part of the at least one candidate second image overlaid by the shifted symbol. The areas over which correlation is performed of the first image and the at least one candidate second image may overlap each other.
The correlation may be performed on at least one of luminance, feature points, texture and colour.
The correlation may comprise at least one of mathematical cross-correlation, mean square error, maximum absolute difference and phase correlation. The step (d) may comprise providing an indication of distance to a position of the camera for capturing the second image.
According to a second aspect of the invention, there is provided an apparatus for performing a method according to the first aspect of the invention.
According to a third aspect of the invention, there is provided an apparatus for selecting a stereoscopic pair of images, comprising:
a display for displaying images;
means for causing the display to display at least one cursor overlying a displayed image;
means for aligning the at least one cursor with a respective image feature of the displayed image;
means for selecting a displayed image; and
means for shifting the at least one cursor laterally by a predetermined amount in response to selection by the selecting means of a first image.
The apparatus may comprise a camera. The display may comprise a camera viewfinder display. The selecting means may comprise a manually operable control for controlling the camera to capture at least the first image.
According to further aspects of the invention, there are provided a computer program for causing an apparatus according to the second or third aspect of the invention and comprising a computer to perform a method according to the first aspect of the invention, a computer program for adapting an apparatus, which comprises a computer and which is not capable of performing a method according to the first aspect of the invention, to perform a method according to the first aspect of the invention, and a carrier medium carrying such a program.
It is thus possible to provide techniques which allow a user to select stereoscopic pairs of images relatively easily and quickly. In the case of capturing images using a camera, this technique guides a user into capturing pairs of images which are substantially rectilinear and which require little or no processing before being displayed by means of a 3D display. It is not necessary for a user to attempt to match two images which are suffering from perspective distortion because of rotation of the camera between capturing the first and second images.
For example, these techniques may be used to guide a person so that they can take two 2D images from good positions so that they are close to being rectilinear. These techniques may be used for guiding a person when taking the second image from a single camera, for selecting the best image as the second image from a sequence of stills or video sequence, or for guiding the adjustment of a dual head camera or two camera system where the separation may be varied.
Structure in the scene is used by the user so that, when the first image is taken, the structure is used as a reference by overlaying a cursor on some image feature. This cursor is translationally shifted and the camera is moved in a parallel manner so that the feature is again under the cursor. The second image is then taken from a parallel position at a regulated camera separation and hence provides controlled depth. A correlation metric may be calculated between the two camera positions and the camera can take the second image automatically when the correlation is at a maximum and the camera is in the correct position.
Position measuring sensors such as GPS (Global Positioning System), compass and tilt sensors may also be used to enhance the system and provide additional feedback to help ensure the two images are as parallel as possible.
The present invention will be further described, by way of example, with reference to the accompanying drawings, in which:
Like reference numerals refer to like parts throughout the drawings.
The optical system 2 comprises a set of lenses and a motor or the like for adjusting the positions of the lenses or lens groups to perform focusing. The camera 1 is of the “autofocus” type and includes an autofocus sensor 6, which supplies to the processor 4 information about the distance from the camera to objects in the scene. The processor 4 controls the autofocus mechanism of the optical system 2 in accordance with information from the sensor 6.
The parts of the camera 1 described so far are essentially conventional. However, the processor 4 is arranged to perform a method of selecting stereoscopic pairs of images as described hereinafter. A conventional shutter release control 7 is provided and is actuated by a user who wishes to capture the image which is currently imaged on the sensor 3 by the optical system 2. The processor 4 is also connected to a “cursor alignment” control 8 which is used to permit manual alignment of a cursor as described hereinafter.
In order to select a stereoscopic pair of images, the method illustrated in the flow diagram of
Alternatively or additionally, the cursor may be aligned automatically with an image feature. For example, the processor 4 may process the signal from the sensor 3 to identify a suitable image feature and may then automatically align the cursor or one of the cursors with that feature on the display 5. In the case where cursors are aligned automatically, the dedicated cursor alignment control 8 may be omitted. Alternatively, cursor alignment may be performed by conventional and existing camera controls which are arranged to permit manual cursor alignment during stereoscopic image capture.
Although the camera is described as being of the digital still type, this method may be used with other types of camera. For example, this method may be used with film cameras having conventional optical viewfinder displays and with additional optical elements for making the cursor visible in the viewfinder display.
When a user is satisfied with the image framing in the display 5 and with the positioning of the or each cursor, such as the rectangular cursor 23, the user takes a first image in a step 13 by actuating the shutter release control 7. The processor stores the image digitally and, in a step 14, automatically moves the position of the cursor on the viewfinder display 5.
The amount of shift may be determined using techniques disclosed, for example, in EP 1 089 573, the contents of which are incorporated herein by reference. By fixing the amount of shift, the amount of depth of an object in a 3D image is fixed irrespective of how close the object is. The autofocus function on the camera 1 may be used to set limits on the amount of depth which is viewable and information from the autofocus system may be used in determining the lateral shift of the cursor.
In the present example, it is assumed that the first captured image is of the view intended for the left eye in a stereoscopic pair. As illustrated in
When the shifted cursor is displayed, the user then moves the camera so as to capture a second of the stereoscopic pair of images. For example, instructions to the user may be displayed on the viewfinder display 5 instructing the user to move the camera 1 to the right until the shifted cursor is correctly aligned on the image feature of the image currently being imaged by the optical system 2 on the sensor 3 and being displayed on the viewfinder display 5.
Although in this example the cursor is described as being displayed for one image then shifted laterally for the second, it is possible to display the cursor in both initial and shifted positions on both images. The user aligns the image feature with the unshifted cursor, takes the first image, and then aligns the feature with the shifted cursor. An indication of the “active” cursor may be provided by highlighting the cursor to be used with, for example, colour or solid/dashed lines.
The moving of the camera and aligning of the cursor are illustrated by a step 15 in
As an alternative or additional technique for capturing the second image, a semi-automatic “one shot” method may be used and is illustrated in
A step 41 determines whether the correlation is above a threshold, which is indicated as being the maximum correlation in this example. Motion estimation theory predicts the general behaviour of a correlation function around the position of a maximum (alternatively, using the phase correlation technique, the maximum is determined directly). As the user moves the camera towards the correctly shifted position, the correlation increases and the maximum can be found by monitoring the change in the value of the correlation. Two techniques which can determine when the maximum has been found to sufficient accuracy are: determining when the correlation starts to decrease; determining when the improvement in correlation falls below a predetermined threshold.
When maximum correlation is detected, a step 42 automatically captures the second image. Thus, the user moves the camera and the second image is automatically captured when the camera is at the position 1′ where correlation is at its maximum.
The step 42 of capturing the second image may also be performed manually by the user, who may utilise a meter indicating the level of correlation to establish a judgement of when to capture the second image. The user takes both images manually, using an indication of correlation to determine an appropriate occasion to capture the second image.
In order for the semi-automatic method to be performed, the area around the cursor 23 must be sufficiently large to provide an accurate measure of correlation. Also, the area should be such as to overlap with its shifted position. If sufficient computing power is available within the processor 4, the area used for measuring correlation may comprise substantially the whole image.
The correlation may be based on luminance information in the images but may alternatively or additionally use other information such as feature points, texture and colour of the images. Any suitable metric may be used to measure correlation and examples of such measures are mean squared error, maximum absolute difference, cross-correlation, and phase correlation for example as disclosed by C. Kuglin and D. Hines in “The Phase Correlation Image Alignment Method”, Proceedings of the IEEE International Conference on Cybernetics and Society, pp 163-165, 1975.
Some of these correlation metrics, such as phase correlation, are capable of indicating the distance to the desired position and this may be used to assist the user in moving the camera to the position from which the second image should be captured. For example, a visual and/or audio indication of the distance from the correct position may be provided by the camera 1. In the case of a visual indication, this may be provided on the viewfinder display 5, for example in the form of a graph or moving bar. In the case of an audio indication, this may be provided by a tone which changes pitch in accordance with distance from the correct position. Phase correlation may be implemented using fast Fourier transforms, which may be performed in real time using acceptable amounts of processing power and are known in the technical field.
Because the first and second images are captured at different times, any moving objects in the scene will be at different locations by the time the camera is moved to the second position and the second image is captured. This causes noise in the 3D effect and looks unnatural. For example, if a 3D image of a person is to be captured, the subject must remain still between capturing the first and second images and this may be difficult. It is therefore advantageous for the second image to be captured quickly after the first image has been captured. The semi-automatic method of second image capture using phase correlation allows the camera to be repositioned quickly so that the delay between capturing the first and second images may be made relatively small.
Although the second image is at a different viewpoint, the correlation will be at a maximum when the camera is purely translated as indicated by the arrow 30 in
It is thus possible to provide a technique which greatly assists users in capturing stereoscopic pairs of images and which requires little or no processing in order to give a comfortably viewed 3D effect when viewed on suitable display equipment. In particular, a user is assisted in shifting the camera by the correct amount and correctly orienting the camera so that rectilinear images are captured.
Although this technique has been described in detail for capturing images with a camera, it may be used for selection of stereoscopic pairs of images from other sources. For example, this technique does not have to be used in “real time” but may be used to select stereoscopic pairs from a video sequence, where a video camera has been translated, or from a sequence of still images which were previously captured. In such examples, a first image may be chosen by inspection or arbitrarily. Other images may then be selected, by inspection or arbitrarily, and the technique may be applied so as to select the “best” stereoscopic pair of images from those available.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US3943344 *||Jun 26, 1974||Mar 9, 1976||Tokyo Shibaura Electric Co., Ltd.||Apparatus for measuring the elevation of a three-dimensional foreground subject|
|US4791478 *||Oct 3, 1985||Dec 13, 1988||Gec Avionics Limited||Position indicating apparatus|
|US5054097 *||Nov 23, 1988||Oct 1, 1991||Schlumberger Technologies, Inc.||Methods and apparatus for alignment of images|
|US5455689 *||Feb 14, 1994||Oct 3, 1995||Eastman Kodak Company||Electronically interpolated integral photography system|
|US5682197 *||May 12, 1995||Oct 28, 1997||Eastman Kodak Company||Electronic panoramic camera for use with an external processor|
|US6225979 *||Jul 14, 1998||May 1, 2001||Sanyo Electric Co., Ltd.||Cursor display method and apparatus for perceiving cursor as stereoscopic image in stereoscopic display region, and recording medium recorded with cursor display program that can be read out by computer|
|US6583814 *||Sep 30, 1999||Jun 24, 2003||Thomson Licensing Sa||System for correction of convergence in a television device related application|
|US6618497 *||Jun 21, 2000||Sep 9, 2003||Pentax Corporation||Photogrammetric image processing apparatus and method|
|US6677981 *||Dec 31, 1999||Jan 13, 2004||Stmicroelectronics, Inc.||Motion play-back of still pictures comprising a panoramic view for simulating perspective|
|US6697573 *||Mar 14, 2001||Feb 24, 2004||Imax Corporation||Hybrid stereoscopic motion picture camera with film and digital sensor|
|US6747610 *||Jul 16, 1998||Jun 8, 2004||Sanyo Electric Co., Ltd.||Stereoscopic image display apparatus capable of selectively displaying desired stereoscopic image|
|US6791598 *||Mar 17, 2000||Sep 14, 2004||International Business Machines Corporation||Methods and apparatus for information capture and steroscopic display of panoramic images|
|US6980690 *||Nov 24, 2000||Dec 27, 2005||Canon Kabushiki Kaisha||Image processing apparatus|
|US7091931 *||Oct 24, 2002||Aug 15, 2006||Geo-Rae Co., Ltd.||Method and system of stereoscopic image display for guiding a viewer's eye motion using a three-dimensional mouse|
|US7123292 *||Sep 29, 1999||Oct 17, 2006||Xerox Corporation||Mosaicing images with an offset lens|
|US7215809 *||Apr 4, 2001||May 8, 2007||Sony Corporation||Three-dimensional image producing method and apparatus therefor|
|US7466336 *||Sep 5, 2002||Dec 16, 2008||Eastman Kodak Company||Camera and method for composing multi-perspective images|
|US20010045979||Jul 26, 2001||Nov 29, 2001||Sanyo Electric Co., Ltd.||Methods for creating an image for a three-dimensional display, for calculating depth information, and for image processing using the depth information|
|US20020001036||Mar 14, 2001||Jan 3, 2002||Naoto Kinjo||Digital camera and image processing method|
|US20020105514 *||Nov 20, 2001||Aug 8, 2002||Roche Michael L.||Stereoscopic image processing on a computer system|
|US20020118454 *||Oct 3, 2001||Aug 29, 2002||John Hey||System for stereoscopic viewing of an image|
|US20020135539 *||Feb 22, 2001||Sep 26, 2002||Blundell Barry George||Interaction with a volumetric display system|
|US20020191841 *||Aug 26, 2002||Dec 19, 2002||Dynamic Digital Depth Research Pty Ltd||Image processing method and apparatus|
|US20030136907 *||Feb 3, 2003||Jul 24, 2003||Hitachi, Ltd.||Charged particle beam apparatus|
|US20030151659 *||Feb 6, 2003||Aug 14, 2003||Pentax Corporation||Camera for generating a stereoscopic pair of images|
|US20030152263||Feb 6, 2003||Aug 14, 2003||Pentax Corporation||Digital camera for taking a stereoscopic pair of images|
|US20040027451 *||Apr 8, 2003||Feb 12, 2004||Image Masters, Inc.||Immersive imaging system|
|US20040041906 *||Sep 4, 2002||Mar 4, 2004||Fruit Theodoric Benjamin||Method for producing a real-time, moving, stereoscopic, preview image|
|GB2370443A||Title not available|
|JP2000305207A||Title not available|
|JP2001230955A||Title not available|
|JP2003244500A||Title not available|
|1||International Search Report for corresponding Application No. PCT/JP2004/051679 mailed Dec. 13, 2004.|
|2||Japanese Office Action for corresponding application No. JP2006-525981 dated Sep. 10, 2009.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US8587518 *||Apr 15, 2011||Nov 19, 2013||Tektronix, Inc.||Disparity cursors for measurement of 3D images|
|US8830231 *||Jun 7, 2011||Sep 9, 2014||Nintendo Co., Ltd.||Storage medium having stored therein stereoscopic image display program, stereoscopic image display device, stereoscopic image display system, and stereoscopic image display method|
|US8842166||Jun 7, 2011||Sep 23, 2014||Nintendo Co., Ltd.||Storage medium having stored therein stereoscopic image display program, stereoscopic image display device, stereoscopic image display system, and stereoscopic image display method|
|US8902298||Jun 7, 2011||Dec 2, 2014||Nintendo Co., Ltd.||Storage medium having stored therein stereoscopic image display program, stereoscopic image display device, stereoscopic image display system, and stereoscopic image display method|
|US9001192||Jun 7, 2011||Apr 7, 2015||Nintendo Co., Ltd.|
|US9098930 *||Sep 27, 2012||Aug 4, 2015||Adobe Systems Incorporated||Stereo-aware image editing|
|US20100091094 *||Jun 1, 2009||Apr 15, 2010||Marek Sekowski||Mechanism for Directing a Three-Dimensional Camera System|
|US20110304615 *||Jun 7, 2011||Dec 15, 2011||Hal Laboratory, Inc.|
|US20120162067 *||Apr 15, 2011||Jun 28, 2012||Tektronix, Inc.||Disparity cursors for measurement of 3d images|
|US20130083021 *||Sep 27, 2012||Apr 4, 2013||Scott D. Cohen||Stereo-Aware Image Editing|
|U.S. Classification||348/218.1, 352/62, 348/50, 382/294, 348/42|
|International Classification||H04N13/02, H04N13/00, H04N5/225, H04N5/232|
|Cooperative Classification||H04N5/23293, H04N13/0022, H04N13/0221, H04N13/021|
|European Classification||H04N5/232V, H04N13/02A1A, H04N13/02A1M, H04N13/00P1D|
|Oct 13, 2006||AS||Assignment|
Owner name: SHARP KABUSHIKI KAISHA, JAPAN
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:HILL, LYNDON;JONES, GRAHAM R.;REEL/FRAME:018384/0420
Effective date: 20060306
|Mar 19, 2015||FPAY||Fee payment|
Year of fee payment: 4